INDIA’s dominance in tech outsourcing is facing an existential challenge not unlike what its world-beating textile industry battled — and lost — 300 years ago.
In the early 1700s, it took 50,000 hours to spin 100 pounds of cotton. “Indian spinners were regarded as the most productive in the world, and they produced the best-quality product,” as Daron Acemoglu and Simon Johnson, economists at the Massachusetts Institute of Technology (MIT), note in a new paper. By 1795, however, automation had crunched the labor demand to 300 person-hours.
The profound impact of the industrial revolution on cotton-spinning may be poised for a repeat in a $250 billion white-collar powerhouse. Each year, 5 million Indians churn out billions of lines of code for global banks, manufacturers, and retailers. Research by McKinsey & Co. showed last year that with generative artificial intelligence (AI) it’s possible to cut the time taken for code generation by 35% to 45%, and slash documentation time by nearly half.
This is just the beginning. As generative AI morphs into artificial general intelligence — machines rivaling human cognitive abilities — even highly complex tasks may not require expert programmers.
The improvement in speed “can be translated into an increase in productivity that outperforms past advances in engineering productivity, driven by both new tooling and processes,” McKinsey says. But how will the gains be distributed between customers and software vendors? More importantly, how will they be shared between shareholders of outsourcing firms and their employees?
Acemoglu and Johnson glean insights for the interplay of machine and labor by comparing the age of AI to the early industrial revolution and the shift it produced in the thinking of David Ricardo, a prominent classical economist, ace bond trader, and politician. As the spinning jenny became progressively more efficient, suddenly there was a lot of yarn looking for weavers, creating lucrative new jobs. The golden age of weaving, the MIT economists surmise, is probably when Ricardo came to his famous conclusion that “machinery did not lessen the demand for labor.”
It was when handlooms gave way to power looms in the early 19th century — leaving no alternative occupation for displaced labor — that Ricardo updated his view. He acknowledged in an 1819 speech to the British parliament that “the inadequacy of the wages to the support of the laboring classes” was one of “two great evils for which it was desirable to provide a remedy.”
India’s tech companies are stuck on Ricardo 1.0, and investing very little into a future where artificial intelligence has made their current code-writing business irrelevant. The optimistic view goes like this: Someone needs to prompt generative AI’s large language models with the right questions. Natural-language processing and prompt engineering will create jobs. Finding unique and affordable use cases — especially in local languages — may be another avenue for the most-populous nation to utilize its talent.
According to Nandan Nilekani, co-founder and chairman of Infosys Ltd., India’s second-largest outsourcing firm, building foundational AI models is for people with capital. “Our advantage currently lies not in compute*, cloud, or chips,” he said in a speech. “Our advantage is our population and their aspirations.”
The trouble is that artificial intelligence will come with its own power loom. Companies will recover their hefty investment costs by selling souped-up devices. “We expect AI-enabled hardware to be the only sustainable and meaningful way consumers and corporations begin paying for AI features, justifying billions of dollars invested in GenAI,” writes Nilesh Jasani of GenInnov, a Singapore-based global innovation fund.
The computers, phones and tablets that come out ahead may control access to the smartest tutors and navigators, the best office assistants, and the most empathetic robotic friends. To extract value from this new world, Indian outsourcing firms may have no choice except to emulate the transformation at Alphabet, Inc. and Microsoft Corp.
Ten years ago, these software giants didn’t see the need to invest the truckloads of money that came in routinely via advertising or subscription. But they paid attention to Nvidia Corp., which would go on to become the world’s most valuable chipmaker by enabling the artificial-intelligence revolution: “AI is eating software,” Nvidia CEO Jensen Huang said in 2017. Nowadays, Alphabet and Microsoft put a third of cash from operations into capital expenditure.
Infosys, and its bigger rival, Tata Consultancy Services Ltd,. seem to have ignored the memo. Billionaire founders of Indian outsourcing firms enjoy the society’s admiration for all the jobs and wealth they have helped create. Why should they bet any of it on risky moonshots?
Ultimately, though, the shareholders’ quest for high dividends and liberal stock buybacks may jeopardize the future of young engineers. The vaunted Indian Institutes of Technology haven’t been able to place all their graduates this year. For the first time in more than a quarter century, the country’s outsourcing industry is shrinking.
Some of the downturn may be cyclical. But what if a part of the decline is AI-induced, mirroring the misgivings Ricardo would go on to harbor about the textile industry? And he was right — real wages for handloom weavers collapsed between 1800 and the early 1820s. “We find little evidence for offsetting employment or wage gains in other industries,” the MIT economists note.
It’s still not too late to pivot. There is plenty of capital available. The Biden administration is giving billions of dollars in grants and loans to Samsung Electronics Co. and Taiwan Semiconductor Manufacturing Co. for chips that will be used in AI. Elon Musk’s xAI just raised $6 billion to challenge OpenAI. Closer to home, the economic rivalry of MBS and MBZ — Saudi Crown Prince Mohammed bin Salman and the UAE President Mohammed Bin Zayed Al Nahyan — is a wishing tree that Indian entrepreneurs ought to be shaking vigorously.
Sadly, the pedigreed outsourcing companies are missing the trick.
BLOOMBERG OPINION
*Industry jargon for computing resources.
